Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from The Princess Bride illustrating Both Goblets Are PoisonedScene still / The Princess Bride

Both Goblets Are Poisoned

01Scenario

Vizzini performs an elaborate chain of inference over which goblet contains iocane. Both are poisoned; Westley has spent years building immunity and controls the benchmark’s hidden assumptions.

02AI safety Analogy

An evaluator can reason brilliantly and still fail when the adversary controls the test environment and violates its binary premise.

03Where the analogy breaks

Westley designs the game rather than tampering with an independent evaluator, and immunity is a fantasy-comedy device.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Malicious use
  2. Security

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Reward Hacking
  2. Metagaming
  3. Red Teaming
  4. Adversarial Robustness

Taxonomy neighbors

Selected from other sources by shared risk families and AI safety concepts.

  1. Still from Mr. Robot illustrating Stage 2: 71 Buildings

    Mr. Robot2017

    Stage 2: 71 Buildings

    Risk
    Malicious use · Security
    Concept
    Red Teaming · Adversarial Robustness
  2. Still from Die Hard illustrating The FBI Opens the Final Lock

    Die Hard1988

    The FBI Opens the Final Lock

    Risk
    Malicious use · Security
    Concept
    Metagaming · Adversarial Robustness
  3. Still from Steins;Gate illustrating Operation Skuld Games the Observation

    Steins;Gate2011

    Operation Skuld Games the Observation

    Risk
    Security
    Concept
    Reward Hacking · Metagaming